1. Field of the Invention
The invention relates in general to a digital-to-analog converter (DAC) and, in particular, to reduce the integral non-linearity (INL) of a DAC.
2. Background of the Invention
Generally, a digital-to-analog converter (DAC) encounters a plurality of non-idealities such as total harmonic distortion (THD), difference nonlinearity (DNL), integral nonlinearity (INL), and etc. To reduce the INL of a DAC, there are many methods proposed. One of the most popular methods is dynamic element matching (DEM).
Although the INL of a DAC is capable of being reduced with a DEM method, numerous problems arise when reducing DEM to practice, such as the implementation of a randomized generator. One substitute method is a pseudo DEM. A pseudo DEM method utilizes a sequence of element arrangement to approach a DEM method.
Please refer to FIG. 1, which shows an embodiment of a first prior art, U.S. Pat. No. 7,679,539, in which a randomized generator is used for deciding the starting point of a plurality of 1-bit DACs. A decoder then chooses a first set in the plurality of 1-bit DACs as the output signal according to of the input digital signals and the starting point. In such a method, however, there may be a glitch in an outputting signal due to a lack of a common 1-bit DAC in two subsequent outputs. For example, a DAC 10 comprises fifteen 1-bit DACs, namely 101˜115. In a first output, the 1-bit DACs 101˜103 are selected. In a second output, the 1-bit DACs 105˜107 are selected. In such an operation, large glitches are more prone to happen.
To avoid such a phenomenon, another PDEM is developed. Please refer to FIG. 2 which illustrates the method of another prior art, US 2011/0,279,292. In this prior art, if a first input signal is 011, the cells 201˜203 are selected. If a second input signal is 101, which is larger than the first input 011, the cells 201˜205 are selected. If a third input signal is 010, which is smaller than the second input 101, the cells 204 and 205 are selected. Hence, there is at least one common cell selected in two subsequent operations. However, in this prior art, it's not taught how to choose cells if the two subsequent inputs have an identical value. As a consequence, the PDEM may fail if the two subsequent inputs have an identical value.
What is needed is a method to reduce the glitch in DEM while preventing the PDEM from failure.